CTO-line gets you delivery velocity but ships engineer-flavoured solutions. COO-line gets you process redesign but underinvests in eval infra. Standalone 'Head of AI' gets you executive air-cover but stalls on integration debt. We argue COO-line under 200 people, CTO-line above 500, with a 12-month checkpoint to detect a wrong placement before it calcifies.
You're hiring (or promoting) the person who will run AI for your company. The question that consumes more board cycles than it should: who do they report to? Three options — under the CTO, under the COO, or standalone reporting to the CEO. Each ships a different kind of AI program. The wrong choice does not break the company; it just produces an AI program that is shaped by reporting line, not by the problem.
CTO-line
AI work ships fast and well-engineered. The rest of the business gets engineer-flavoured solutions: tools instead of workflows, APIs instead of operations. Eval infrastructure is solid. Change management is uneven — engineers ship the model, then ask the business to figure out how to use it. Best for companies whose bottleneck is platform: distributed systems, integration surface, infra cost discipline.
COO-line
AI work shows up as process redesign first, technology second. The team thinks in workflows, KPIs, and adoption curves. Eval infrastructure tends to underinvest because nobody on the COO's side knows how much they need. Engineering velocity is uneven — the AI team is asking favours from a real engineering team that has its own roadmap. Best for companies whose bottleneck is operational discipline: ops-heavy, regulated, or workflow-rich.
Standalone (Head of AI reporting to CEO)
Executive air cover is highest. The AI program looks credible to the board, gets resources, gets a seat at the strategy meeting. Failure mode: integration debt. The Head of AI ships pilots that are great in isolation but cannot connect to anything because they have no engineering authority and no operational authority. Best for companies where AI is the strategic story, not a productivity multiplier — funded startups where the next round depends on it.
COO-line under 200 people. CTO-line above 500. Standalone only when AI is the strategic story.
The 12-month checkpoint
Whatever placement you pick, audit it at the 12-month mark. Three signals that the placement is wrong:
- Eval infrastructure is missing or skeleton-only. Symptom of COO-line that did not invest in measurement.
- Adoption is below 30% on shipped tools. Symptom of CTO-line that shipped without change management.
- Pilots stuck in pilot. Symptom of standalone with no integration authority.
If you see any of these, the fix is rarely to re-org. It is to staff the gap: a delivery lead under a CTO-line AI team; an eval engineer under a COO-line AI team; an integration engineer embedded with the standalone Head of AI. The reporting line is half the answer; the staffing is the other half.
A practical rule
For companies under 200 people: COO-line, because the bottleneck at that size is process, not platform. For companies above 500: CTO-line, because the bottleneck shifts to platform. In between, it depends on whether your business is more ops-heavy or more engineering-heavy. If you cannot tell, default to COO-line and add an engineering lead inside the team.